Florida Interlock Program Driver Characteristics: Factors Predictive of SelfSelection

1
Florida Interlock Program Driver
CharacteristicsFactors Predictive of
Self-Selection Compliance
Robert B. Voas A. Scott Tippetts Milton Grosz
2
2 Primary Questions

• Among those who are eligible, what type of driver
  chooses to install an interlock? (26)
• Of this subset (those who install an interlock),
  what factors predict a higher likelihood of
  compliance and completion? (82.4)

3
Types of Predictive Factors
4
Analytic Methods

• 3 Logistic Regressions (3 outcome variables)
• of Total Eligible - Installed Interlock vs.
  Didnt Install
• of Installed subgroup - BAC Test Violation vs.
  Not
• of Installed subgroup - Dropped Out vs. Not
• Forward Conditional Variable Selection
• (only include/retain factors
  significant at plt.05)
• 2 models for each regression
• predicting from Demographics Driving Record
  only
• same as above, plus residential SES (imputed
  from ZipCode)

5
Question 1 Who Chooses to Get an Interlock
Installed? by Age
6
Question 1 Who Chooses to Get an Interlock
Installed? by Sex
7
Question 1 Who Chooses to Get an Interlock
Installed? by Ethnicity
8
Question 1 Who Chooses to Get an Interlock
Installed? by Prior Offenses
Percent of Eligible Drivers Installing Interlock,
by Priors
9
Other Factors Prior Driving Record / Licensing
Procedural Compliance
10
Question 1, model 2 still predicting
Self-Selection for Interlock Installation, but
with imputed SES measures factored in

• College Graduates (p.010)
• Unemployment Rate (p.015)
• Low Income Households lt25k / Poverty Rates
  (p.011)
• NOTABLE CHANGES IN OTHER PREDICTORS
• Sex effect no longer significant (p.156)
• Ethnicity effect dampened for Blacks and
  Hispanics
• (from 30 less likely, to 24 less likely)
• Ethnicity effect augmented slightly for Asians
• (from 72 more likely, to 81 more likely)

11
Question 2a What factors predict Interlock
Violations?

• DEMOGRAPHICS
• Sex not significant (p.150)
• Age younger more likely, older less
  (p.027)
• (general monotonic relationship)
• Blacks 23.6 more likely than whites (p.035)
• Hispanics 55.8 less likely than whites
  (lt.001)
• PRIOR DRIVING RECORD
• Prior DUIs 6.6 less likely per prior
  (p.011)
• Prior Refusals 16.0 less likely per prior
  (plt.001)
• Habitual Traffic Offender 34.8 more likely
  (p.016)

12
Question 2a What factors predict Interlock
Violations? with SES

• SES of Drivers Neighborhood (imputed for group)
• More High School Graduates in zipcode more
  likely to violate
• Higher Median Income of zipcode more likely to
  violate
• Higher Employment rate in zipcode more likely
  to violate
• Higher proportion urban area less likely to
  violate
• DEMOGRAPHICS
• Age effect is no longer significant (Sex still
  not)
• Blacks no longer any more likely than whites
• Hispanic effect (55.8 less likely) remains
  unchanged
• PRIOR DRIVING RECORD no change
• Prior DUIs 6.6 less likely per prior
  (p.011)
• Prior Refusals 16.0 less likely per prior
  (plt.001)
• Habitual Traffic Offender 34.8 more likely
  (p.016)

13
Question 2b What factors predict Interlock
Drop-Out? (includes some violators)

• DEMOGRAPHICS
• Sex not significant (p.253)
• Age above 50 years less likely (p.018)
• Blacks 47.4 more likely than whites (p.034)
• Hispanics no effect (plt.656)
• PRIOR DRIVING RECORD
• (Prior Refusals, HTOs not significant this time)
• Prior DUIs 20.1 more likely per prior
  (plt.001)
• Failure to Pay Fines / Tickets 35.3 more
  likely (p.002)
• Failure to Appear (Summons) 68.3 more likely
  (p.044)
• NOTE inclusion of SES factors makes no
  difference

14
Conclusions Implications

• Other studies of recidivism risk / correlates
• prior driving offenses of all types
• males 30-49 years ethnic groups
• SES factors, which partially correlate with
  demographics
• Those who install Interlocks are different, and
  likely have a lower a priori risk of recidivism
  to begin with
• Those who do go on Interlock, but then violate,
  fail to comply, or drop-out are also different
  from those who complete, and are likely at a
  particularly high risk of future recidivism
• Implications for Screening / Promotion /
  Mandating?
• Implications for Evaluation Analysis (e.g.,
  comparing post-Interlock Recidivism with
  non-Interlock offenders)?